systems engineering//cyber-physical system
A cyber-physical system is a system in which software and physics share one closed loop: code reads sensors, decides and drives actuators, and the physical world answers back through the same sensors, so that a computation error becomes a wrong movement rather than a wrong number on a screen. The term is used to name the class of systems (drones, bottling lines, trains with automatic braking, water treatment plants, a power grid regulating its frequency) whose software must meet demands office software never meets: real-time deadlines, stability and physical safety.
A cyber-physical system is a system in which software and physics share one closed loop: code reads sensors, decides and drives actuators, and the physical world answers back through the same sensors, so that a computation error becomes a wrong movement rather than a wrong number on a screen. The term is used to name the class of systems (drones, bottling lines, trains with automatic braking, water treatment plants, a power grid regulating its frequency) whose software must meet demands office software never meets: real-time deadlines, stability and physical safety.
The structure is the one every closed-loop system shares: a state that evolves under the software's action and the world's disturbances, sensors that report part of it with noise, and a policy that computes the next action from what was measured. What makes the system cyber-physical is that the policy is code on a processor, with its own timing, failure modes and attack surface, and that the state is mass, heat, pressure or voltage. Every bug therefore has a physical signature: the drone's attitude estimate arriving 30 ms late shows up as oscillation, not as an error message.
In a CPS, software errors have mass and momentum. A late computation is a phase lag, a crash is an actuator frozen at its last command, a malicious packet can open a valve; so timing, failure behaviour and security are design requirements of the software, verified with the same rigour as its logic.
A CPS is often confused with the Internet of Things. In IoT the centre is connecting devices and collecting data, and the loop, if any, closes slowly and often through a person; in a CPS the centre is the loop and its guarantees. A sensor sending a temperature to the cloud is IoT; the thermostat switching the boiler is a CPS, if a modest one.
Every robot is a CPS, but not every CPS is a robot: a power grid has no arms. The drone is the textbook case because it is unstable without control, its sensors are cheap and noisy and it decides in milliseconds (drone).
Safety and security meet here. An attack on the software can end as a physical accident, which is the subject of cyber-physical security and the reason security requirements now sit beside safety ones.
The cost of thinking in CPS terms (timing analysis, hazard analysis, fault behaviour) pays whenever a software error can break something physical. Software that only fills a monthly report is an information system with different rules.
Building one that demonstrably works is systems engineering; the computer it runs on is an embedded system.